THE RECOMMENDATION LAYER · MAY 18, 2026 · 9 MIN READ

Net New Intelligence: The Only Content AI Has a Reason to Quote

AI absorbs generic content without credit and cites what adds to the record. Net New Intelligence, the one-question test, and how a small firm produces it.

By Jax Baker
TL;DR — WHAT TO REMEMBER

Net New Intelligence is our name for content that gives AI engines something they cannot get anywhere else: your original data, your named method, your first-hand experience, your claimed positions. Engines absorb generic content without credit because they already learned it from thousands of pages just like it; they cite what adds to the record. If your library could have been written without you, it will be read without citing you.

You have been publishing for years. Blog posts, tips, explainers, the monthly newsletter your web person said Google likes. The writing is competent, the advice is sound, and the engines have given you nothing for any of it. No citations, no mentions, no buyers arriving saying “the AI recommended you.”

Here is the uncomfortable explanation: the engines already knew everything your last twenty posts said. They learned it from the ten thousand posts just like them. When your article tells a machine what it already knows, the machine absorbs the confirmation and moves on. Nothing about that transaction requires your name.

At Probably Genius we call the content that breaks this pattern Net New Intelligence, and it is the quality bar behind everything we publish and build. This guide covers why good content gets absorbed without credit, what the engines demonstrably quote instead, the one-question test to run on your own library, and what producing net-new material actually looks like for a small firm.

Why Your Best Posts Get Read and Never Credited

An AI engine building an answer is doing a job with two very different inputs. Most of what it reads is confirmation: the same advice, the same definitions, the same seven tips, repeated across thousands of pages with the nouns swapped. A small fraction is information: something it could not have assembled from the rest.

Confirmation is useful to the machine. It just isn’t creditable. When five hundred estate-planning blogs all explain what a revocable trust is, the five-hundred-and-first explanation makes nothing better, and the engine has no reason to attach anyone’s name to knowledge that belongs to the whole pile.

Our co-founder Jacquie Baker published the sharpest version of this mechanism in a February 2026 essay, and her image is worth carrying whole: “Your original expertise is a high-resolution photograph. Sharp. Detailed. Unique.” Republishing the consensus, she wrote, is copying the copies: “You can’t photocopy a photocopy and expect it to be worth citing.”

That is the whole diagnosis of the library that never earned a mention. The posts were not bad. They were photocopies, and the engine had the original stack.

This is not a new idea the AI era invented, either. Google published a patent in 2020 for scoring documents by their “information gain”: how much new, non-redundant information a page adds relative to what the reader has already seen. The machines have been trying to price novelty for years. The answer engines just raised the stakes, because now the price of redundancy is total: absorbed, used, unnamed.

What the Engines Have a Reason to Quote

The good news is that what earns citations is not a mystery. It has been measured, repeatedly, and the findings agree.

The foundational academic study on generative engine optimization, presented at KDD 2024 across a benchmark of 10,000 queries, found that adding citations to credible sources, quotations, and statistics to a page boosted its visibility in generative engine responses by up to 40 percent, with specific statistics among the strongest single levers. Not longer content. Not more content. Content carrying verifiable, specific substance.

Semrush’s January 2026 study of 11,882 prompts across ChatGPT Search, Google AI Mode, and Perplexity found the same shape from the commercial side: citation correlated most strongly with clear summarization (+32.83 percent), strong E-E-A-T signals (+30.64 percent), question-and-answer formatting (+25.45 percent), and structured data (+21.60 percent). E-E-A-T is the machine asking: does this page demonstrate first-hand experience and identifiable expertise, or could anyone have typed it?

Format follows intent, too. A March 2026 analysis of 75,000 AI answers found engines matching the type of page they cite to the type of question asked, with ranked lists taking 40.9 percent of commercial-intent citations. The engines are not citing at random and they are not citing volume. They cite the page that adds something checkable to the answer they are assembling.

Put the studies side by side and the pattern reads like a shopping list: original numbers, named sources, demonstrated experience, extractable structure. Every item on that list is something the ten-thousand-post pile cannot fake, which is exactly why it is what gets quoted.

The Test: Could the Machine Have Written It Without You?

You do not need a study to audit your own library. You need one question, and Jacquie published it in the same February 2026 essay: “Could ChatGPT write this without talking to me?”

Her scoring is blunt: “If yes: low Information Gain. Invisible. If no: high Information Gain. Citable.”

Run it on your five most recent posts, honestly. The seven-warning-signs listicle: could the machine have written it without you? Almost certainly, which means it already has, thousands of times. The what-is-X explainer? Same. Now find the paragraph where you mentioned what a client engagement actually costs, or the case that taught you to stop trusting a common industry assumption, or the number of times a certain problem crossed your desk last year. The machine could not have written that sentence. That sentence is the asset.

The libraries we look at tend to be mostly photocopy, with the occasional original sentence buried mid-post, unclaimed and unstructured, where no engine will ever surface it. The problem was never that the firm had nothing new to say. It is that the new thing was seasoning, when it should have been the meal.

Where Your Net New Actually Lives

Owners hear “original data and claimed positions” and picture a research department. Wrong picture. For a working firm, Net New Intelligence is lying around in four places, undocumented.

Your numbers. What things cost, how long they take, how often each outcome happens across your real caseload. You quote these on the phone every week. Nobody has published them, including you.

Your method. The sequence you actually follow, the checks you refuse to skip, the order of operations you learned the hard way. If it has no name and no page, it reads to a machine like everyone else’s process. Named and documented, it is citable intellectual property with your entity attached.

Your war stories. The client who almost made the expensive mistake, what it would have cost, what you did instead. First-hand experience is the E in E-E-A-T for a reason: it is the one input the content mills structurally cannot supply.

Your positions. The advice you give in the first meeting that your competitors hedge on. The industry default you think is wrong, and why. A claimed position is risky at a dinner party and priceless on the record, because a machine assembling an answer needs someone who actually said the thing.

Notice that all four already exist. This is the part the content-marketing industry keeps getting backwards: the firms with the least AI visibility are often the ones sitting on the most uncopyable material, because they spent their years in the trenches doing the work while the people the machines cite were busy getting published.

You are not short on material. You are short on record.

How a Small Firm Produces It

Producing Net New Intelligence is a discipline, and it runs in three moves.

Claim it. Put your name on the method, the numbers, the positions. Unclaimed expertise gets credited to whoever claimed something similar first, and the engines are attribution machines: AI can’t cite what you won’t claim. Information gain is what claiming produces.

Structure it. One canonical page per claim, written so the core answer can be lifted whole: direct definition first, evidence underneath, your name and entity wired through it. This is the “quotable answers” rung of the Recommendation Layer, and it is where a strong claim either becomes machine-readable or evaporates into brochure copy.

Back it. Real figures with sources, dated and checkable. A page that makes claims it can support survives the engines’ corroboration checks; a page that rounds up dies by them. We hold our own content to a written standard on exactly this point, and we publish the standard: every substantive piece carries verifiable sources, and every number has somewhere to point.

The honest caveat, because this doctrine gets oversold elsewhere: net-new content does not guarantee a citation, this week or any week. Engines weigh many signals, and visibility is measured in percentages, not promises. What the evidence does support is the asymmetry: commodity content has approximately zero chance of being credited, and content only you could have written is the entire class of material the engines have a reason to quote. You already own the raw ingredient. Getting it claimed, structured, and backed is translation work, and translation is our whole job.

Want to Learn More?

Probably Genius was built by Jacquie Baker and Christopher Shaw, who spent more than 40 combined years translating what makes an expert the best in the room for the audiences that decide who gets chosen. Net New Intelligence is the standard that work answers to: if a page could exist without the client, it does not deserve the client’s name, and we do not ship it.

Frequently asked questions

What is Net New Intelligence?
Net New Intelligence is Probably Genius’s term for content that adds something to the record AI engines cannot get anywhere else: original data from your own practice, a named and documented method, first-hand experience, and publicly claimed positions. It is the opposite of commodity content, which engines absorb without attribution because they already learned it from thousands of similar pages. Net New Intelligence is the class of material engines have a reason to cite and attribute.
Why doesn’t AI cite my blog posts?
Most business blogs restate advice the engines have already seen thousands of times, and a machine has no reason to credit anyone for knowledge that exists identically across the whole category. Research on generative engines shows citations concentrate on pages carrying verifiable specifics: statistics, quotations, credible sources, and demonstrated first-hand experience. If a post could have been written by anyone in your industry, it gets read as confirmation and absorbed without a name attached.
What kind of content do AI engines actually cite?
Measured studies point the same direction: content with specific statistics, quoted sources, and citations gained up to 40 percent visibility in generative engine responses in the foundational KDD 2024 benchmark, and a January 2026 Semrush study of 11,882 prompts found clear summarization, E-E-A-T signals, Q&A formatting, and structured data most associated with getting cited. In plain terms: original numbers, named methods, first-hand experience, and extractable structure earn citations. Volume and polish alone do not.
How do I add information gain to my content?
Start with the test: could an AI engine have written the piece without you? If yes, add what it could not know. Publish your real numbers (costs, timelines, outcome frequencies from your own caseload), name and document your method on a canonical page, tell the first-hand stories only your practice produced, and state the positions you actually hold. Then structure each claim so the core answer can be quoted directly, and back every figure with a checkable source.

CITATIONS

  1. “GEO: Generative Engine Optimization” (Aggarwal et al., KDD 2024). The foundational academic study of generative engine visibility across a 10,000-query benchmark, finding that adding citations, quotations, and statistics boosted content visibility in generative engine responses by up to 40 percent. dl.acm.org
  2. “Contextual Estimation of Link Information Gain” (United States Patent US20200349181A1, published November 2020). Google’s patent describing a system that scores and ranks candidate documents by the volume of new, non-redundant information they provide relative to documents the user has already seen. patents.google.com
  3. “How We Built a Content Optimization Tool for AI Search” (Semrush, January 2026). Correlation study of 11,882 prompts across ChatGPT Search, Google AI Mode and Perplexity: clear summarization (+32.83 percent), E-E-A-T signals (+30.64 percent), Q&A formatting (+25.45 percent) and structured data (+21.60 percent) were most associated with getting cited. semrush.com
  4. “AI citations favor listicles, articles, product pages: Study” (Search Engine Land, March 2026). Analysis of 75,000 AI answers across ChatGPT, Google, and Perplexity finding citation formats align with query intent, with listicles, articles and product pages driving 52 percent of citations and ranked lists taking 40.9 percent of commercial-intent citations. searchengineland.com
  5. “The Moat AI Can’t Cross – You” (Jacquie Baker, Jax5d, February 2026). The essay by our co-founder quoted above: the photocopy argument and the Information Gain self-test. jax5d.substack.com
WRITTEN BYJacquie ("Jax") Baker

Founder of Probably Genius, an AI visibility firm helping professional service brands become the named answer in AI search. Nearly two decades across technology, digital strategy and branding — including 1,000+ digital projects through her previous agency — now focused on making experts visible, verifiable and recommendable to AI. Let's talk →

Related reading

Read the method.
Then see it run on you.

The 109-point diagnostic is this library, applied to your business. Free, about an hour to present, no obligation.

See where you stand →

A real diagnostic, not a sales call in disguise.